Research On Anomaly Detection Of Operation And Maintenance Data Based On Deep Learning

In the information age,there are a large number of Internet service companies that need to closely monitor a large number of network service data to ensure the stable operation of their businesses.This is what internet companies call operation and maintenance now.However,for data with different patterns and varying qualities,simple human observation can be quite labor-intensive and far less accurate than expected.With the development of artificial intelligence,operation and maintenance have entered the era of intelligent operation and maintenance accordingly.Anomaly detection is a basic and important function of Artificial Intelligence for IT Operations(AIOps)system,which aims to automatically detect abnormal fluctuations in operation and maintenance data through algorithms,and provide corresponding decision-making basis for subsequent abnormal warning and root cause analysis.In the actual scenario,due to the scarcity of abnormal point data,different data modes and diversity of data types,the detection of operational and maintenance data anomalies has brought great challenges.First of all,this paper focuses on the operational data of the company,an insurance industry in view of the actual scene of the actual demand,combined with deep learning technology,respectively,using the LSTM,VAE and GAN network solving data anomaly detection in the intelligent operations,and methods of this article is to provide unsupervised learning method,has solved the most existing operational data anomaly detection for supervised learning method requires a large number of manpower and time for the problem of annotation data.Secondly,this paper also integrates two anomaly detection methods of operation and maintenance data,data abnormal prediction methods and multidimensional correlation analysis,and proposes a set of intelligent operation and maintenance system solutions.At present,the system is on line in an insurance company and has been recognized by the enterprise.Finally,this paper analyzes and compares the performance differences of all the methods proposed in this paper,and compares the operation and maintenance data anomaly detection method based on deep learning with the existing operation and maintenance data anomaly detection method through experimental comparison.The anomaly detection method of operation and maintenance data based on deep learning has better detection effect than the existing methods on some types of operation and maintenance data and is similar to the existing methods on some types of operation and maintenance data.